Scene graph prediction has emerged as a critical task in computer vision,focusing on transforming complex visual scenes into structured representations by identifying objects,their attributes,and the relationships amo...Scene graph prediction has emerged as a critical task in computer vision,focusing on transforming complex visual scenes into structured representations by identifying objects,their attributes,and the relationships among them.Extending this to 3D semantic scene graph(3DSSG)prediction introduces an additional layer of complexity because it requires the processing of point-cloud data to accurately capture the spatial and volumetric characteristics of a scene.A significant challenge in 3DSSG is the long-tailed distribution of object and relationship labels,causing certain classes to be severely underrepresented and suboptimal performance in these rare categories.To address this,we proposed a fusion prototypical network(FPN),which combines the strengths of conventional neural networks for 3DSSG with a Prototypical Network.The former are known for their ability to handle complex scene graph predictions while the latter excels in few-shot learning scenarios.By leveraging this fusion,our approach enhances the overall prediction accuracy and substantially improves the handling of underrepresented labels.Through extensive experiments using the 3DSSG dataset,we demonstrated that the FPN achieves state-of-the-art performance in 3D scene graph prediction as a single model and effectively mitigates the impact of the long-tailed distribution,providing a more balanced and comprehensive understanding of complex 3D environments.展开更多
In this paper, a novel component-based scene graph is proposed, in which all objects in the scene are classified to different entities, and a scene can be represented as a hierarchical graph composed of the instances ...In this paper, a novel component-based scene graph is proposed, in which all objects in the scene are classified to different entities, and a scene can be represented as a hierarchical graph composed of the instances of entities. Each entity contains basic data and its operations which are encapsulated into the entity component. The entity possesses certain behaviours which are responses to rules and interaction defined by the high-level application. Such behaviours can be described by script or behaviours model. The component-based scene graph in the paper is more abstractive and high-level than traditional scene graphs. The contents of a scene could be extended flexibly by adding new entities and new entity components, and behaviour modification can be obtained by modifying the model components or behaviour scripts. Its robustness and efficiency are verified by many examples implemented in the Virtual Scenario developed by Peking University.展开更多
Background In this study,we propose a novel 3D scene graph prediction approach for scene understanding from point clouds.Methods It can automatically organize the entities of a scene in a graph,where objects are nodes...Background In this study,we propose a novel 3D scene graph prediction approach for scene understanding from point clouds.Methods It can automatically organize the entities of a scene in a graph,where objects are nodes and their relationships are modeled as edges.More specifically,we employ the DGCNN to capture the features of objects and their relationships in the scene.A Graph Attention Network(GAT)is introduced to exploit latent features obtained from the initial estimation to further refine the object arrangement in the graph structure.A one loss function modified from cross entropy with a variable weight is proposed to solve the multi-category problem in the prediction of object and predicate.Results Experiments reveal that the proposed approach performs favorably against the state-of-the-art methods in terms of predicate classification and relationship prediction and achieves comparable performance on object classification prediction.Conclusions The 3D scene graph prediction approach can form an abstract description of the scene space from point clouds.展开更多
Scene graph is a infrastructure of the virtual reality system to organize the virtual scene with abstraction, it can provide facility for the rendering engine and should be integrated effectively on demand into a real...Scene graph is a infrastructure of the virtual reality system to organize the virtual scene with abstraction, it can provide facility for the rendering engine and should be integrated effectively on demand into a real-time system, where a large quantities of scene objects and resources can be manipulated and managed with high flexibility and reliability. We present a new scheme of multiple scene graphs to accommodate the features of rendering engine and distributed systems. Based upon that, some other functions, e.g. block query, interactive editing, permission management, instance response, ＂redo＂ and ＂undo＂, are implemented to satisfy various requirements. At the same time, our design has compatibility to popular C/S architecture with good concurrent performance. Above all, it is convenient to be used for further development. The results of experiments including responding time demonstrate its good performance.展开更多
Scene graphs of point clouds help to understand object-level relationships in the 3D space.Most graph generation methods work on 2D structured data,which cannot be used for the 3D unstructured point cloud data.Existin...Scene graphs of point clouds help to understand object-level relationships in the 3D space.Most graph generation methods work on 2D structured data,which cannot be used for the 3D unstructured point cloud data.Existing point-cloud-based methods generate the scene graph with an additional graph structure that needs labor-intensive manual annotation.To address these problems,we explore a method to convert the point clouds into structured data and generate graphs without given structures.Specifically,we cluster points with similar augmented features into groups and establish their relationships,resulting in an initial structural representation of the point cloud.Besides,we propose a Dynamic Graph Generation Network(DGGN)to judge the semantic labels of targets of different granularity.It dynamically splits and merges point groups,resulting in a scene graph with high precision.Experiments show that our methods outperform other baseline methods.They output reliable graphs describing the object-level relationships without additional manual labeled data.展开更多
An ultra-massive distributed virtual environment generally consists of ultra-massive terrain data and a large quantity of objects and their attribute data, such as 2D/3D geometric models, audio/video, images, vectors,...An ultra-massive distributed virtual environment generally consists of ultra-massive terrain data and a large quantity of objects and their attribute data, such as 2D/3D geometric models, audio/video, images, vectors, characteristics, etc. In this paper, we propose a novel method for constructing distributed scene graphs with high extensibility. This method can support high concurrent interaction of clients and implement various tasks such as editing, querying, accessing and motion controlling. Some application experiments are performed to demonstrate its efficiency and soundness.展开更多
针对基于关系边界框提取的谓词特征粒度相对较粗的问题,提出区域敏感的场景图生成(region-sensitive scene graph generation,RS-SGG)方法。谓词特征提取模块将关系边界框分为4个区域,基于自注意力机制抑制关系边界框中与关系分类无关...针对基于关系边界框提取的谓词特征粒度相对较粗的问题,提出区域敏感的场景图生成(region-sensitive scene graph generation,RS-SGG)方法。谓词特征提取模块将关系边界框分为4个区域,基于自注意力机制抑制关系边界框中与关系分类无关的背景区域。关系特征解码器在进行关系预测时不仅考虑了物体对的视觉特征和语义特征,也考虑了物体对的位置特征。在视觉基因组(visual genome,VG)数据集上分别计算了RS-SGG方法针对场景图生成、场景图分类和谓词分类3个子任务的图约束召回率和无图约束召回率,并与主流的场景图生成方法进行了比较。实验结果表明,RS-SGG的图约束召回率和无图约束召回率均优于主流方法。此外,可视化实验结果也进一步证明了所提出方法的有效性。展开更多
基金supported by the Glocal University 30 Project Fund of Gyeongsang National University in 2025.
文摘Scene graph prediction has emerged as a critical task in computer vision,focusing on transforming complex visual scenes into structured representations by identifying objects,their attributes,and the relationships among them.Extending this to 3D semantic scene graph(3DSSG)prediction introduces an additional layer of complexity because it requires the processing of point-cloud data to accurately capture the spatial and volumetric characteristics of a scene.A significant challenge in 3DSSG is the long-tailed distribution of object and relationship labels,causing certain classes to be severely underrepresented and suboptimal performance in these rare categories.To address this,we proposed a fusion prototypical network(FPN),which combines the strengths of conventional neural networks for 3DSSG with a Prototypical Network.The former are known for their ability to handle complex scene graph predictions while the latter excels in few-shot learning scenarios.By leveraging this fusion,our approach enhances the overall prediction accuracy and substantially improves the handling of underrepresented labels.Through extensive experiments using the 3DSSG dataset,we demonstrated that the FPN achieves state-of-the-art performance in 3D scene graph prediction as a single model and effectively mitigates the impact of the long-tailed distribution,providing a more balanced and comprehensive understanding of complex 3D environments.
基金Project supported by the National Basic Research Program (973) of China (No. 2004CB719403), and the National Natural Science Foun-dation of China (Nos. 60573151 and 60473100)
文摘In this paper, a novel component-based scene graph is proposed, in which all objects in the scene are classified to different entities, and a scene can be represented as a hierarchical graph composed of the instances of entities. Each entity contains basic data and its operations which are encapsulated into the entity component. The entity possesses certain behaviours which are responses to rules and interaction defined by the high-level application. Such behaviours can be described by script or behaviours model. The component-based scene graph in the paper is more abstractive and high-level than traditional scene graphs. The contents of a scene could be extended flexibly by adding new entities and new entity components, and behaviour modification can be obtained by modifying the model components or behaviour scripts. Its robustness and efficiency are verified by many examples implemented in the Virtual Scenario developed by Peking University.
基金Supported by National Natural Science Foundation of China(61872024)National Key R&D Program of China under Grant(2018YFB2100603).
文摘Background In this study,we propose a novel 3D scene graph prediction approach for scene understanding from point clouds.Methods It can automatically organize the entities of a scene in a graph,where objects are nodes and their relationships are modeled as edges.More specifically,we employ the DGCNN to capture the features of objects and their relationships in the scene.A Graph Attention Network(GAT)is introduced to exploit latent features obtained from the initial estimation to further refine the object arrangement in the graph structure.A one loss function modified from cross entropy with a variable weight is proposed to solve the multi-category problem in the prediction of object and predicate.Results Experiments reveal that the proposed approach performs favorably against the state-of-the-art methods in terms of predicate classification and relationship prediction and achieves comparable performance on object classification prediction.Conclusions The 3D scene graph prediction approach can form an abstract description of the scene space from point clouds.
基金Supported by National Natural Science Foundation of China(Nos.61173080,61232014,61472010,61421062)National Key Technology Support Program of China(No.2013BAK03B07)
文摘Scene graph is a infrastructure of the virtual reality system to organize the virtual scene with abstraction, it can provide facility for the rendering engine and should be integrated effectively on demand into a real-time system, where a large quantities of scene objects and resources can be manipulated and managed with high flexibility and reliability. We present a new scheme of multiple scene graphs to accommodate the features of rendering engine and distributed systems. Based upon that, some other functions, e.g. block query, interactive editing, permission management, instance response, ＂redo＂ and ＂undo＂, are implemented to satisfy various requirements. At the same time, our design has compatibility to popular C/S architecture with good concurrent performance. Above all, it is convenient to be used for further development. The results of experiments including responding time demonstrate its good performance.
基金This work was supported by the National Natural Science Foundation of China(Nos.62173045 and 61673192)the Fundamental Research Funds for the Central Universities(No.2020XD-A04-2)the BUPT Excellent PhD Students Foundation(No.CX2021222).
文摘Scene graphs of point clouds help to understand object-level relationships in the 3D space.Most graph generation methods work on 2D structured data,which cannot be used for the 3D unstructured point cloud data.Existing point-cloud-based methods generate the scene graph with an additional graph structure that needs labor-intensive manual annotation.To address these problems,we explore a method to convert the point clouds into structured data and generate graphs without given structures.Specifically,we cluster points with similar augmented features into groups and establish their relationships,resulting in an initial structural representation of the point cloud.Besides,we propose a Dynamic Graph Generation Network(DGGN)to judge the semantic labels of targets of different granularity.It dynamically splits and merges point groups,resulting in a scene graph with high precision.Experiments show that our methods outperform other baseline methods.They output reliable graphs describing the object-level relationships without additional manual labeled data.
基金Supported by the National Basic Research Program of China (Grant No. 2004CB719403)the National High-Tech Research & Development Program of China (Grant Nos. 2006AA01Z334, 2007AA01Z318, 2009AA01Z324)+1 种基金the National Natural Science Foundation of China (GrantNos. 60573151,60703062,60833007)the Marine 908-03-01-10 Project
文摘An ultra-massive distributed virtual environment generally consists of ultra-massive terrain data and a large quantity of objects and their attribute data, such as 2D/3D geometric models, audio/video, images, vectors, characteristics, etc. In this paper, we propose a novel method for constructing distributed scene graphs with high extensibility. This method can support high concurrent interaction of clients and implement various tasks such as editing, querying, accessing and motion controlling. Some application experiments are performed to demonstrate its efficiency and soundness.
文摘针对基于关系边界框提取的谓词特征粒度相对较粗的问题,提出区域敏感的场景图生成(region-sensitive scene graph generation,RS-SGG)方法。谓词特征提取模块将关系边界框分为4个区域,基于自注意力机制抑制关系边界框中与关系分类无关的背景区域。关系特征解码器在进行关系预测时不仅考虑了物体对的视觉特征和语义特征,也考虑了物体对的位置特征。在视觉基因组(visual genome,VG)数据集上分别计算了RS-SGG方法针对场景图生成、场景图分类和谓词分类3个子任务的图约束召回率和无图约束召回率,并与主流的场景图生成方法进行了比较。实验结果表明,RS-SGG的图约束召回率和无图约束召回率均优于主流方法。此外,可视化实验结果也进一步证明了所提出方法的有效性。
